作者位次 #9/11ACL 2026Findings· Findings of ACL
APEX : Learning Adaptive Priorities for Multi-Objective Alignment in Vision-Language Generation
Dongliang Chen · Xinlin Zhuang · Junjie Xu · Luojian Xie · Zehui Wang · Jiaxi Zhuang · Haolin Yang · Liang Dou · Xiao He · Xingjiao Wu · Ying Qian
摘要
Multi-objective alignment for text-to-image generation is commonly implemented via static linear scalarization, but fixed weights often fail under heterogeneous rewards, leading to optimization imbalance where models overfit high-variance, high-responsiveness objectives (e.g., OCR) while under-op...